Mike Heffner

Mike Heffner

@mike.heffner.io

AI Mechanic - Cursor Previously: Rotel.dev @ Streamfold, Netlify, SolarWinds Cloud, Librato. Just add water.

57 followers81 following57 posts

Hello future GitHubber! Know that this HTML response will be vomited upon all failures regardless of client, API, or request type surface from now until eternity. So keep it tight.

Serverless Lambda on managed servers

AWS News Feed on 🦋@awsrecentnews.bsky.social

🆕 AWS Lambda now supports up to 32 GB memory and 16 vCPUs for Managed Instances, ideal for compute-intensive workloads. Customers can choose memory-to-vCPU ratios (2:1, 4:1, 8:1) for optimal resource allocation. Available in all regions with Lambda Managed Instances. #AWS #AwsLambda

aws.amazon.comAWS Lambda supports up to 32 GB of memory and 16 vCPUs for Lambda Managed InstancesAWS Lambda now supports up to 32 GB of memory and 16 vCPUs for functions running on Lambda Managed Instances, enabling customers to run compute-intensive workloads such as large-scale data processing, media transcoding, and scientific simulations without managing any infrastructure. Customers can also configure the memory-to-vCPU ratio — 2:1, 4:1, or 8:1 — to match the resource profile of their workload. Lambda Managed Instances lets you run Lambda functions on managed Amazon EC2 instances with built-in routing, load balancing, and auto-scaling, giving you access to specialized compute configurations including the latest-generation processors and high-bandwidth networking, with no operational overhead. Customers building compute-intensive applications such as data processing pipelines, high-throughput API backends, and batch computation workloads require substantial memory and CPU resources to process large datasets, serve low-latency responses at scale, and run complex computations efficiently. Previously, function execution environments on Lambda Managed Instances were limited to 10 GB of memory and approximately 6 vCPUs, with no option to customize the memory-to-vCPU ratio. Functions on Lambda Managed Instances can now be configured with up to 32 GB of memory, and a choice of memory-to-vCPU ratio — 2:1, 4:1, or 8:1 — allowing customers to select the right balance of memory and compute for their workload. For example, at 32 GB of memory, customers can configure 16 vCPUs (2:1), 8 vCPUs (4:1), or 4 vCPUs (8:1) depending on whether their workload is CPU-intensive or memory-intensive. This feature is available in all AWS Regions where Lambda Managed Instances is generally available. You can configure these settings using the AWS Console, AWS CLI, AWS CloudFormation, AWS CDK, or AWS SAM. To learn more, visit the AWS Lambda Managed Instances product page and documentation.

Backends implemented on split ingestion pipelines are in for a rough time

OpenTelemetry@opentelemetry.io

‼️ #OpenTelemetry is deprecating the Span Events API. Our aim is to remove the confusion caused by having overlapping ways to emit events: span events and log-based events. Read our latest post to learn more about how you can prepare. buff.ly/0hOwfuk

Mike Heffner reposted

You can test new tech ideas using the Seinfeld Test Would the product eliminate the plot of an episode? (Google maps, cell phones, paypal, battery packs) Good tech. Would the product inspire new Seinfeld plots? (NFTs, AI chatbots, crypto currency, blindboxes, metaverse land sales) Bad tech.

Are there any known solutions to setting up Claude Code authentication against a subscription plan in a headless, ephemeral environment?

Thanks to ClickHouse for inviting us to share this post. Had a lot of fun digging into performance benchmarking work again!

ClickHouse@clickhouse.com

Can you 4x OTel thoughput with Rust? Meet Rotel: pushing 460K spans/sec/core (vs 130k) How? By diving deep into Rust's threading model, resolving allocator lock contention, and reengineering JSON serialization with RowBinary encoding to stream data efficiently into ClickHouse.

The glue in many of these cases are the bindings built on PyO3. We found it powerful enough to allow us to go from Rust->Python->Rust when building native Python processor integration for Rotel.

The New Stack@thenewstack.io

Rust usage for Python extensions surged 22% in one year as developers choose memory safety and C-level performance over traditional approaches. By @darrylktaft.bsky.social

When you've got to use up the garden tomatoes that are about to go bad: sandwish.netlify.app/ac70ab777318...

sandwish.netlify.appToo Many TomatoesA truly delicious creation. Stack yours and share it with the world!
Eduardo Bouças@eduardoboucas.com

Sandwish is a fun little app that lets you build your signature sandwich and share it with the world. Here's my masterpiece, The Full Stacker: sandwish.netlify.app/b953c4891950... Go check it out and create your own! If you're interested in what's inside the kitchen, read on 🧵

The past 7 months @ramond.bsky.social and I have worked on the hypothesis that telemetry data requires a systems approach for high-perf, predictable and resource efficient collection at scale. We're excited to announce our progress so far and hope that the community can join us: dub.sh/A7sI0Tm

dub.shRotel: Fast and Efficient OpenTelemetry Collection in Rust | RotelObservability shouldn't be a resource hog, that’s why at the end of last year, we started an ambitious new project – Rotel. We’ve been excited to see the rise of OpenTelemetry as a vendor neutral stan...

Mike Heffner reposted

Excellent article on positioning domain engineering between product engineering and infrastructure engineering. I fully agree with the financial model section. However, I prefer to call it ‘platform engineering: the hard parts,’ simply because in the middle, we face more volatile semantics.

jackdanger.comInfrastructure Gravity & Domain Engineering | Jack DangerThe following is an excerpt from Executive Engineering.   “You become responsible, forever, for what you have tamed.” – Antoine de Saint-Exupéry, in The Little Prince   Each company draws its own ...

Another reason, past cold start times, I'm glad we focused on optimizing our extension initialization time. It would be great if SecretsManager API calls were faster, since that's an often the long pole in the tent for init time. github.com/streamfold/r...

github.comGitHub - streamfold/rotel-lambda-extension: AWS Lamba Extension of the Lightweight Rotel CollectorAWS Lamba Extension of the Lightweight Rotel Collector - streamfold/rotel-lambda-extension

Mike Heffner reposted

I wanted to write something like this so many times when I was at Netlify! We always decided it was best to not make a fuss. Now I'm not at Netlify, I still believe abstracting the hosting layer via adapters makes for better framework architecture overall, as well as being better for the ecosystem.

netlify.comNext.js Deployment Challenges: Why platforms need better open source collaborationDiscover the 6 key challenges platforms face when supporting Next.js deployments, including the lack of adapter support and undocumented behaviors. Learn how Netlify is working with OpenNext to build ...

I've definitely found this to be true and trying to debug it via traditional means like Google and SO is impossible since no human would have made that same error.

Simon Willison@simonwillison.net

Here's the table of contents for my lengthy new piece on how I use LLMs to help me write code simonwillison.net/2025/Mar/11/...

Mike Heffner reposted

BIG NEWS. I have left the Post. Corporate and billionaire media are failing to meet the moment. With @normeisen.bsky.social I'm launching @contrariannews.bsky.social. We'll have politics but also cooking, humor, film and even pets. Please subscribe and join the fight. contrarian.substack.com

contrarian.substack.comThe Contrarian | SubstackUnflinching journalism in defense of democracy. Click to read The Contrarian, a Substack publication. Launched 13 hours ago.

The ordered sequence numbers provided in the metadata would seem to provide enough to build an S3 log streaming Warpstream/Bufstream equivalent without an external consistency store. However, at 10 min lag you wouldn't be able to stream anything in real time.